Living Jackson

Benefits of cycling
How To Test Your Bike To Make It Faster | GCN Does Science

How To Test Your Bike To Make It Faster | GCN Does Science

(soothing music) – So you want to go faster on your bike and you’re looking for aero gains. For example, a time trial handlebar like this Time Zone bar from ControlTech, or maybe new wheels, or an aero helmet, or a new position? But how do you know if
it’s actually faster rather than just feeling faster? Well here at GCN we like
to talk about science. We even sometimes do proper experiments and oh how we love to quote peer reviewed research from scientific journals. But the problem is that the results we get in experiments might not be relevant to you. It’s normally very case specific, especially for aerodynamics. If you want to test what’s really faster for you then you’d need to go to a wind tunnel, which is quite expensive. Or you could do a really good field test. So here is GCN’s guide to how to do a really good field test to test for aero improvements. First of all, let’s talk
about input variables and response variables. That’s kind of a fancy way of saying that you need to change something and measure something. So the thing you’re changing is what is going to hopefully bring you an aerodynamic improvement and the thing you’re measuring is probably going to be whether you go faster or not. So in our example what I’m going to change is the angle on these ControlTech aero bars from zero degrees, where they are at the
moment, to five degrees. And that might sound
like a tiny tiny change but it actually is the kind of thing that we would’ve measured in a wind tunnel the year before in Olympic Games with British Cycling to see if it brought any benefits. In fact, we would’ve measured even smaller changes than that. Now the test I’m going to use is a time trial effort on this one kilometer
stretch of straight road and the response variable is going to be whether I go any faster. The key things you want
to pay attention to when designing an experiment are validity, repeatability, reliability, and measurement. Or should that be measureability? Internal experimental validity means quite simply making sure that you are measuring
what you want to measure and not something else instead. It requires you to make
sure your experiment is consistent so you just have to think quite carefully about other factors that might influence the output apart from the different input variables, like in my case the bar angle. So right now I’m thinking about all the possible reasons that I might go faster on my time trial stretch of road that aren’t due to being more aerodynamic. For example, if one time I put down 50 watts more than another time then I’ll probably go
faster but it won’t be due to changing the angle of the bars. So basically what you need to do is think about keeping
all the variables constant apart from the one variable
that you’re changing. And this is also known as local control. Local control means trying to remove all other influences on your experiment beyond the one variable you are deliberately changing. For example, the variables you don’t want to change are, wind conditions and temperature, so do all your tests on the same day if possible and in calm conditions. The power you exert, traffic, fatigue, tire pressure, body position, the total weight of your bike, the total weight of you and your clothing, and so on. So to put it another way, you only want to change one variable at that time and that does include weight. So if for example I was to go and have lunch mid experiment, well that would change my weight and that would influence the results. So lunch stops, no coffee stops and for that matter no pee stops either mid experiment and if the variable you are changing on your bike changes the weight of the bike, for example, if you go from Lo-Pro wheels to deep section wheels then you need to compensate for that weight difference. For example, what you can is have a bead on for both setups but with more water for the lower weight setup of your bike. So for my experiment today I’ve chosen a route
with no traffic lights, no corners and not too
much traffic either. There is a bit of wind but it’s not very gusty and it’s fairly consistent. Of course I have a power meter so I can make sure that my average power of each time trial
test run is the same. Now if you don’t have a power meter it is still possible to do
an aero test like this. What you need is a very gradual decent where you can just freewheel down and measure the difference in time. But it is better with a power meter because the movement of your legs actually affects your
aerodynamic resistance. Another key thing is I’m not going to ride this test full gas. It’s not just because I’m lazy it’s also because I’ve got to do it at least six times and talk more about repeatability later. Although I do have to do it fairly fast to make the test externally valid. External validity is
another important concept. It means the extent to which the results of your experiment can be assumed to be true for other cases. For example, different people, different courses or different times. Well, now when it comes to riding a bike and specifically the
effect of aerodynamics we know from science that some differences are person and shape specific, i.e. sadly what is faster for you might not be faster for someone else or on a different course. That’s kind of obvious
if you think about it. So what this means is that basically your experiment is unlikely to be externally valid but that doesn’t actually really matter because you want to find
out if you’re faster, not if anyone is faster. What it does mean though is that you need to choose a test course similar to the course
that you’re targeting or racing, whether it’s rolling, flat, hilly, windy, calm, make sure it’s as similar as possible. Choosing the right speed is also really important for aerodynamic testing. That’s because your aerodynamic resistance increases as you go faster so if you do your test too slowly you might not see any difference even though you might actually have improved your aerodynamic drag. Now we come to repeatability. For each angle of the aero bars I’m going to repeat the test and do it three times. Why three times? Well, so that I can see any anomalies. For each setup the results should be fairly similar but fluky results or outliers are always a possibility, especially when you’re
testing out on the road. So if I do it three times, or even better would be five times but we’re a bit short of time, then I can see if one result is really different from others and that gives me a clue that something might have gone wrong with that test and I can repeat it. It is important to note though that you should only discount results that are actually outliers, not just the results that you
don’t like the look of. Reliability might sound like repeatability but it’s actually subtlely different. Reliability is about how much of an effect the variable you’re changing has compared to the inevitable effect of outside influences and of course how accurately you can measure things. For example, I’m testing
two different bar angles and I’m doing three test
runs for each angle. Now if there’s more a difference between the times for
the same angle of bar, i.e. the same setup than it is between the averages for different angles that means my test is not very reliable, or to put it another way, the difference between the two setups is smaller than I can actually measure so you probably don’t need to worry about bar angle in that case. Now it’s going to sound blindingly obvious but you need to be able to measure your output variable, so in this case the time taken for each test run. Now I need to press lap at exactly the same point at the start and finish of each lap. That way I can take the data file home and analyze it later and see what the numbers really said and I don’t have to work out averages while I’m riding along on the road. While we’re talking
measurement your choice of time trial test route
is extremely important. It needs to be similar
to your race course, yes, but not the same length. It should be shorter because remember you’re going to have to repeat it three times for each variable. On the other hand it also shouldn’t be too short because otherwise the slight differences when you push the lap button on your bike computer will lead to quite a significant error compared to the total length of the time that you’re measuring. (soothing music) So here we are back at GCN HQ and it’s time to analyze the results of my aero bar angle experiment. So results are in. It was actually a 1.1 kilometer course I’d like to point out. For zero degrees I averaged 212.33 watts and the average time was
1 minute 47.67 seconds and for five degree bar angle I averaged 213 watts and the average time was 1 minute 48.33 seconds. And looking at those
numbers one might assume that zero degrees is faster
but hold your horses. I made some real mistakes designing this experiment, for example, there’s more variation between the time measured for each setup. So for example at zero
degrees the time measured varied from 1 minute 43 seconds to 1 minute 51 seconds
and for five degrees the quickest time was 1 minute 44 seconds and the slowest time
was 1 minute 55 seconds. So what you can see is there’s more variation in time within the setup, so for the same setup, then there is between the two setups. So if there is a difference it’s so small that we’d have to do a statistically significant number of tests to work it out and three is definitely not statistically significant. Secondly that large
variation within each setup is probably due to the gusty wind. So I could really feel the wind blowing me around when I was out there and like I said earlier you want to avoid the influence of wind on this kind of test. Well, I hope this video helps you in your quest
for aero improvements. Give it a thumbs up if so and share it with your
speedy cycling friends. Don’t forget to let me know down in the comments if you’re setting up your own aero experiment and if you’d like to check out another aero geekry video there’s this one about why size really is important.

100 comments on “How To Test Your Bike To Make It Faster | GCN Does Science

  1. Nice video about the scientific method. One thing I would like to add: You should first be sure about all influencing variables, ideally with a rough idea how much they influence the measured values. for example you won't care about changes of solar winds if you measure speeds with different handle bar angles 🙂
    In your case I would say that the most important influence are the wind conditions, far more important than even changes of a few grams of weight or even the handle bar angle. So unless you get that completely under controll, you won't be able to isolate the influence of the changed variable.
    But that won't matter anyway, because it just means that since you cant controll that external wind variable, you wont be able to controll it in real life either and it will always be "as unpredictable" as it has been during measuring. So the solution would be to run more like 100 test runs, to cancel out that wind-variations over time and get the average benefit of your changed handle bar angle.

  2. One thing I noticed was your elbow pads were not under your elbows. Shouldn't they be a bit further back ? I don't do that kind of test, I find my time is better invested in getting fitter which is probably the most significant way most people can get faster. A bike fit is a very good investment though. But once you get your position dialed in I think it's unlikely that any further adjustment will lead to significant time improvements. I would love to try one of those devices that give you a realtime drag coefficient value. It would allow one to make incremental changes while not having to go through such a tedious process.

  3. How refreshing too see and hear a clear and thorough explanation of the basics of experimental design. Thank you Emma.

  4. Wouldve loved you see if Velocomp's aeropod was used , when used with a direct force power meter it can measure CdA numbers.

  5. I expected the title to be "thinking while cycling". I cycle as stress relief and sometimes I find clarity and then (eventually) great ideas pops in my mind. NVM im not english speaker myself so that's why I was misled. I enjoyed the content anyway.

  6. Good job we had Emma doing this one – I'd imagine an entirely different set up if Dan was let given a free reign to experiment with "bar angles"….

  7. I suspect that to do enough repetitions to have adequate statistical power to pull a signal out of the noise you would need a longer course and a quite large number of trials, necessitating several days of testing. Try to do it all in one day and fatigue confounds. But one would have to make sure both variables are tested the same number of trials each day. My guess is it is pretty hard to ever get a convincing result without a wind tunnel.

  8. Joking aside its really refreshing to have such a video that brings the lessons of testing from experience as a top elite racer into a format that would be useful for us all. Also good to see it not providing a great deal of useful info on improving her position with such a minor change, because if it had done so then it would have posed large questions about the efficacy of the testing Emma had done as an elite rider!

  9. Looks like you were up on Bannerdown, Emma. It is quite exposed to the elements there; perhaps somewhere more sheltered might have been better. Incidentally, I would like to see more of your 'oat cuisine' recipes. Thank you.

  10. I don't know anything about aerobars but I do know about biomechanics and I suspect, on balance, the usefulness of aerobars is extremely specific and limited. So much about putting ones arms in that position just seems wrong–even more so for anyone larger than Emma.

  11. I have been thinking to replace the spoke reflectors with reflective tape on rims, but I doubt it will give much difference for me… By traffic law, they count the reflective stripe on tire as legal as spoke reflectors. So I guess it could be on rim too.. when using disc brakes. But it's snowing now, so I don't think I would be able to do good science anytime soon. ^_^

  12. I always find it funny when I see those people riding the bike path on their tri bars and I pass them like they were having a recovery ride or something.

  13. Nice vid. Just a couple of things to add:
    1. Find a road with a steady gradient that puts the wind as much as possible from behind. Wind has a lot less influence from behind than from any other angle. This can mean that you have to wait for a day with the right conditions.
    The thing with tail wind is, though, that it is far more difficult to ride at a stable power output with tail wind than with head wind, which brings me to:
    2. There is a lot of concentration and time involved in doing this kind of testing. Speed and power should be as stable as possible within each effort (normalized power and average power should idealy be exactly the same). If power and speed vary a lot, a) conditions may be too variable to perform a good test, b) the road you selected may not be best suited for testing, or c) you may not be experienced enough to perform these kinds of tests. Riding with a really stable power output over several intervals requires quite a bit of riding experience.
    3. As soon as you see the conditions changing, you should abort. If you can see circumstances are different, they are already way too different to come to any real conclusion.
    4. You should do the runs with alternating setups (run 1 with 0deg, run 2 with 5deg and so on). Small changes in circumstances are than smoothed out better. Yes, this is extra hassle, but it helps smooth out smaller changes like rising temperatures.

  14. It would seem that a 3K, or even a 5 K TT, with a 500m run up to speed, would give you more accurate results. An ideal location would be an outdoor velodrome that has "open" training days.

  15. How to do science on your bike? Easy: just use that moments off-line and without office/lab distraction to think about your projects. That safes tons of workload time down the road. Just try it!

  16. Nice video Emma. Good explanation of the key principles to follow and consider when doing this kind of experimental testing. If you’re not already familiar with it, I suggest checking out Robert Chung’s virtual elevation method. I’ve been using it for the last few months to determine CdA and Crr changes, and I’ve found it to be more effective than other field test protocols that I’ve tried in the past.

  17. Emma, great video. How accurate is Strava for measurement of such an experiment? Maybe you could check with Strava how long of a Segment I need to overcome the measurement error.

  18. Annoyingly, the only way to accurately measure the aerodynamic differences between small changes on bike setup is to use a wind tunnel – having spent hours wind tunnel testing bike setups for my dissertation!

    Great explanation of designing experiments though 🙂

  19. You could eliminate the variation in time (speed) and power by dividing mph(kph) by power. A more aero setup would give you greater speed/watt. Then holding power or speed within a small amount of variation could give you some valid information from a small number of repeats.

  20. This is why GCN needs Emma, or someone with her real science background, contributing. A truly impressive explanation of experimental science concepts applied to the real cycling world.

  21. I enjoyed this and Emma's presentation style is great.My only comment is that given the effect of the variability of the wind on the test results would it be better to do this test on a turbo ?

  22. Cool video. One addition: any confounding variables that are measurable can be included in the statistical model. There are plenty of multivariate techniques available that allow you to build a richer model with more than a single variable, the most obvious and intuitive example probably being a linear regression analysis.

  23. Emma, as usual, did a great job of making the science understandable. As a scientist myself, I can appreciate her skills.

  24. This is where a Strava segment can be useful. Set one then ride through it the amount of times you want to do the test. That way you don't have to worry about hitting start and stop on your computer each time, Just make sure to hit the start of each segment at the same speed / wattage.

  25. use standard deviation because that will tell you about distribution of the results which you need to know because the outliers in the stats are skewing your averages. Its a phenomenon called base rate neglect.

  26. you were using big ring with big cog 😵 you could trash the derailleur like that😂 is that a crossed chain test?

  27. What about just using Load the app onto your Garmin and connect your phone to your Garmin and you are good to go, Cda on the fly. Easy and simple.

  28. @ There are so many factors making these tests inaccurate. I would encourage GCN to do a test with It is also a "on the road" test app that you can download to your Garmin and I belive it is free?, compare it to the manual tests that you do and show us the results. GCN Science 🙂

  29. Great vid as ever Emma, we use exactly the same techniques for continuous improvement. changing uncontrollable to controllable inputs using correlation and regression metrics to identify which are the most important, never really thought about applying that to cycling!

  30. Nice video! Concerning the experimental design, it would probably better to do your replicates not in a row but alternating them. Otherwise you are more fresh in the beginning than in the end (although you should see that when recording power)

  31. lol…pointless for me….10 year old bike…I don't think going faster is an issue…I either will or wont….don't think im going to start putting aero bars n stuff on it…would be worth more than the bike…

  32. Great clip Emma and GCN! Yes I tested today make the same ride as I did in beginning of April for see if my average
    pulse have going down. The wind was the problem as Emma said in the video and I had a little more song wind today I think but from the results I understand my average pulse had go down 10 beats at minute. Next time will be more easy compere since I have a power meter.

  33. Very nice. A dot plot had probably helped to illustrate that mean values can be misleading is there is large variability between the different data points (runs). Sorry geek at work…

  34. Compensating for weight is wrong, you cant randomly assign a weight to a wheel. The weight of a wheel is an inherent property of the wheel you are testing. A field test is good for this because the differences between wheels are not purely aero. To get even more into the weeds tires on that wheel are even harder to control, unless both wheels have the same internal and external rim width, and you are using the same tires and tubes at the same pressures between the two wheels. But if the wheels have different internal and external rim widths, that becomes another inherent variable to the different wheels.

  35. I use this website to control for my differences in speed and power between each test, and yield a result in CdA. You have to do this because if you go faster it requires more watts in a non linear fashion of course.

  36. Hi as a o.a.p I do find this interesting but I remember the days as a young lad on a Sunday morning getting out of bed running down the stairs
    Jumping on my bike riding to a club meeting witch was a 30min ride going in the cafe having a cup of tea and a slice of toast then going out all
    day with the lads never went back to the cafe with them as it was in a different location never had all the technology like we do today
    I still love going out on my bike when I can

  37. How fast is fast enough to matter? Just for example, I cruise about 30kmh, so would I benefit more from positioning for power, or aero positioning? Thanks in advance.

  38. Feeling faster is better than being faster , you have a better mindset making you want to ride more. Witch equals better training and better speed .

  39. Man, how the channel has grown. Depending on the discipline, Emma can look resplendent in her World and British bands where poor old Matt had electrical tape. Interesting topic too…

  40. Not only should you try to avoid having to factor in the wind, but you also need to make sure that whatever wind there is, you need to make that a headwind, as any tailwind would make you faster the less aero your bar angle is.

  41. Excellent primer on experimental design.  But where is the mathematic formula for the statistical outlier test? You've got to show some formulas, Emma!   🙂

  42. AERO FORKS – everything else is screwed as youre on it ! this is why pinarrello forks look as they do and sod the rest..

  43. Emma as usual this was a well informed video. In reality with all the variables, folks interested in a test of this sort should save up for a wind tunnel test. Thanks again.

  44. So, let's see: Emma is a world class athlete, with a PhD in engineering end here she is brilliantly explaining how the scientific method works.
    Her existence means that there's hope for humanity.

    By the way: what she says applies not only to aero tests for cyclists, but to everyone who wants to set foot in a lab as well. Great content, keep it coming!

  45. Nice job Emma. What I would like to see is a GCN attempt at using Aerolab in Golden Cheetah. This video shows how hard it is to do aero testing using the obvious time/power metrics. You could actually make it more amusing by taking Si and seeing which one of you can do a better job. My money is definitely on you.

  46. Isn’t it better to just make a workout-file on like Garmin Connect or so. There you can set the length of your intervals (time or distance). So for this experiment I’d recommend to set it by length because there will always be Some kind of difference when you have to press the lap button every time

  47. If there is no real time difference in the test one is carrying out, choose the position which you find the most comfortable.

  48. I would think that maintaining a constant speed on a given course (preferably circular) for each test and then comparing wattage required to maintain given speed for each change would be an easy way to run these test.

Leave a Reply

Your email address will not be published. Required fields are marked *